The metrics that lie
| Metric | What it rewards | Why it misleads |
|---|---|---|
| Hours studied | Time at a desk | Says nothing about what happened during them |
| Topics covered | Turning pages | Covered isn't learned; a skim ticks the same box |
| Pages of notes | Transcription | The most copied notes come from the least understanding |
| Videos watched | Playback | Recognition without any retrieval at all |
| Streak length | Showing up | Useful for habit, useless as a measure of knowledge |
None of these is worthless — hours and streaks are genuinely useful for building the habit. The mistake is treating them as measures of learning, because that's what makes a productive-feeling term end in a disappointing mark.
The metrics that don't
- Questions answered correctly, cold. The closest thing to a direct measurement. You cannot fake it by sitting still longer.
- Percentage of the syllabus you can explain without notes. Sample five topics, try to explain each, count how many you get through. Brutal and accurate.
- Past-paper marks under timed conditions. The only metric measured in the same units as the outcome you care about.
- Retention on second attempt. Getting something right today means little; getting it right a week later without seeing it in between means a lot.
- Error-log shrinkage. The number of recurring mistakes that have stopped recurring — see how to learn from mistakes.
The common property: each requires retrieval to move. That's the test for whether a metric is worth tracking at all.
Track coverage and depth separately
One number can't represent revision progress, because two different things go wrong. You can have seen everything shallowly, or three topics deeply and the rest not at all — and those need opposite responses.
- 1
List every topic on the syllabus in one column
From the module handbook, not from your notes. Your notes reflect what you attended.
- 2
Rate each on a three-point scale
Can't do it / can do it with notes / can do it cold under time. Three states is enough; five invites false precision.
- 3
Rate honestly by testing, not by feeling
One question per topic, answered cold. Self-rating without testing produces a chart of your confidence, which is not the same chart.
- 4
Re-rate weekly, in the same sitting each time
The movement is the signal. A topic that slid from green to amber in a fortnight is information you can act on.
- 5
Let the amber column set next week's plan
The point of a tracker is to decide what to do next. If it doesn't change the plan, it's decoration.
Leading and lagging indicators
Past-paper marks are a lagging indicator: accurate, but they only move after weeks of work, which makes them useless for steering day to day. Daily retrieval accuracy is a leading indicator: noisier, but responsive enough to tell you whether this week is working.
Track one of each. Daily cards or questions answered correctly gives you the fast signal; a timed past paper every week or two gives you the honest one. Using only the fast signal makes you overconfident, and using only the slow one means you find out too late to change anything.
Keep the tracking cheap
- Under two minutes a day. Elaborate trackers become the hobby, and a beautifully-maintained dashboard is procrastination with a spreadsheet on it.
- One page, visible. A tracker you have to open won't be opened by week four.
- Automatic where possible. Review software already records accuracy and retention; there's no reason to copy those numbers by hand.
- Weekly review, not daily analysis. Look at the numbers once a week in the same slot — see the weekly review. Daily inspection is noise.
What to do when the numbers are bad
The point of measurement is to change something, and there are only really three responses: change the method, change the allocation, or accept the position and re-plan around it.
If retrieval accuracy is flat while hours are high, the method is wrong — usually re-reading rather than testing. If one subject is red and the others green, it's allocation. If everything is red with two weeks to go, it's triage: pick what you can save. What you must not do is respond to bad numbers by stopping measuring, which is the most common response and the reason people arrive at exams without knowing.
The one number to check weekly
If you keep nothing else: what percentage of the syllabus can you currently explain, cold, to the standard the exam requires? Sample five topics at random each week and score out of five.
It's crude, it takes twenty minutes, and it correlates with your final mark far better than anything else you could log. Everything else on this page is refinement — see how to set study goals for turning that number into next week's plan.